BrowseComp
1,266 hard-to-find questions with short answers for testing how persistently agents can search the web.
Reference by OpenAI
About
Use cases: llm-evaluation, web-search
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
About the benchmark
- Benchmark site
- arxiv.org/abs/2504.12516
- Maintainer
- OpenAIarxiv.org
- Launched
- 2025-04arxiv.org
We link to the official leaderboard instead of copying scores. Benchmarks get gamed and saturate; we never use them to rank listings.
Signals
- Stars
- 4,673
- Forks
- 517
- Open issues
- 41
- Last commit
- Apr 22, 2026 (6 months ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- Repository
- github.com/openai/simple-evals
- Website
- arxiv.org/abs/2504.12516
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
More from OpenAI
- 0.708score
security-review-openai
OpenAI
Perform language and framework specific security best-practice reviews and suggest improvements.
SkillBFreshness: B★ 895Claude Code +9 - 0.708score
xlsx-processing-openai
OpenAI
Toolkit for comprehensive Spreadsheet reading, creation, editing, and analysis with visual quality control.
SkillBFreshness: B★ 895Claude Code +9 - 0.708score
docx-processing-openai
OpenAI
Toolkit for comprehensive document reading, editing, and creation with visual quality control.
SkillBFreshness: B★ 895Claude Code +9 - 0.704score
skill-creator-openai
OpenAI
Guide for creating effective skills.
SkillBFreshness: B★ 895Claude Code +9 - 0.703score
pdf-processing-openai
OpenAI
Toolkit for comprehensive PDF reading, reviwing, and creation with visual quality control.
SkillBFreshness: B★ 895Claude Code +9 - 0.695score
OpenAI Codex
OpenAI
OpenAI's coding agent for the terminal, editor, desktop app and ChatGPT cloud, included in ChatGPT plans
AgentAFreshness: A★ 130.8K↓ 13.1M/wk
Cite this page
BrowseComp benchmark. Agent Mart. https://agentmart.dev/reference/browsecomp (last updated 2026-10-24)